# print(heatmap)
# np.savetxt("timepoints_sub5_18reps.csv", heatmap, delimiter=',')
n_comp = 4
csp = CSP(n_components=n_comp, reg=None, log=True, norm_trace=False)
#x, y = data.load_all(cut=[start, end])
#y = np.where(y==1)[1]
#print(x.shape)
print(timepoints)
x, y = data.load_single([5], cut=False)
x = x[0]
y = np.where(y[0] == 1)[1]
for i in range(steps):
csp.fit_transform(x[:, :, timepoints[i]:timepoints[i] + p_size], y)
for j in range(n_comp):
layout = read_layout('EEG1005')
csp.plot_patterns(info,
colorbar=False,
layout=layout,
ch_type='eeg',
size=1.5,
show=False,
show_names=False,
components=j,
title="{}".format(int(
(timepoints[i] - 768) / 0.512)))
#plt.savefig('fig/sub9_time_{}_comp_{}.png'.format(i, j))
clf = Pipeline([('CSP', csp), ('MLP', mlp)])
#clf.set_params(CSP__reg=0.5)
for i in range(100):
shuffle(labels)
scores = cross_val_score(clf,
epochs_data_train,
labels,
cv=10,
n_jobs=1)
accs[sub - 1 if sub < 10 else sub - 2, i] = np.mean(scores)
# Printing the results
# print(labels)
# plot CSP patterns estimated on full data for visualization
csp.fit_transform(epochs_data, labels)
layout = read_layout('EEG1005')
csp.plot_patterns(info,
colorbar=True,
layout=layout,
ch_type='eeg',
size=1.5,
show=False,
show_names=False,
scalings=1,
title="4 first CSP components")
plt.savefig('fig/csp_4comp_sub5_cbar.png')
np.savetxt("csp_mlp_scores.csv", accs, delimiter=',')
